Prognostic utility of SOX2, STAT3, and CD44high/CD24low expression in penile cancer

Penile cancer has a high incidence in developing countries. The standard treatment is removal of the primary tumor and, when necessary, inguinal lymphadenectomy. Currently, the most important prognostic factor is lymph node disease, however, the available staging methods are inaccurate, and the high morbidity rate of lymphadenectomy has stimulated the study of predictive biomarkers of lymph node metastasis for selecting the patients who need lymphadenectomy. SOX2, STAT3 and CD44high/CD24low were chosen because they have provided good predictive results in other squamous cell carcinoma (SCC), although there are no studies for penile cancer. Thus, the expression of SOX2, STAT3, CD24+, and CD44+ in the penile cancer tumor microenvironment was investigated for correlation with tumor behavior in SCC. This observational, prospective, translational study included 34 men and investigated the expression of SOX2, STAT3, CD24+, and CD44+ in tumor tissue by flow cytometry. The median age of the 38 evaluated patients with penile cancer was 61 (37–80) years. Most patients presented a tumor located on the glans penis (82.3%), with the usual histological type (79.4%) and 61.7% of patients presented stage pT2. No metastasis was found in 85.3% of patients. The expression of SOX2, STAT3 and CD44high/CD24low in the microenvironment of penile SCC treated with lymphadenectomy was significantly associated with aggressive tumor behavior (p < 0.05). STAT3 expression shows discrepant points when evaluated in context of angiolymphatic vascular invasion. SOX2, STAT3 and CD44high/CD24low in penile SCC can be indicators of prognosis, allowing for selection of more aggressive treatment when necessary.


Introduction
A total of 1640 new penile cancer cases were diagnosed in the United States in 2014 (Reis et al. 2010).
A systematic review by Montes Cardona and García-Perdomo (2017) showed a global incidence of 0.84/100,000 penile cancer cases annually. The highest incidence was described in Latin America with 1.4 cases/100,000 men, and among Latin American countries, Brazil had the highest incidence of penile cancer, 5.7/100,000 men-year, followed by Paraguay with an incidence of 4.44/100,000 men. In Brazil, penile cancer is a serious public health problem, affecting approximately 2.1% of all male neoplasms. The North and Northeast regions hold 53% of all the cancer cases in the country. In the state of Maranhão, the incidence of the disease was 6.1/100,000 men, and in Pernambuco, particularly, the incidence of penile cancer was 2.1/100,000 men, similar to the national rates (Fonseca et al. 2010;Couto et al. 2014;Coelho et al. 2018). The following conditions are considered risk factors for penile SCC: chronic sebaceous secretions on the glans of the penis, termed smegma, phimosis, human papillomavirus infection, chronic balanoposthitis, lichen sclerosus, and xerotic balanitis obliterans.
In addition, ultraviolet phototherapy, smoking, sexual history with multiple partners, and sexual activity with animals (bestiality) also predispose individuals to SCC (Zequi et al. 2012;Montes Cardona and García-Perdomo 2017).
Tumor invasion and metastatic spread are the main obstacles in the control of SCC. The process of metastasis begins at an early stage in the development of the primary tumor. This hypothesis is reinforced by the observation that, for a tumor to exceed a diameter of 1 or 2 mm, the process of neovascularization or angiogenesis must occur; this phenomenon is an important factor for the development of metastasis (Jing and Zheng 2010). The acquisition of a metastatic phenotype probably requires additional genetic alterations to the genes involved in the tumorigenesis process itself, including increased expression of oncogenes and/or downregulation of tumor suppressor genes, promoting invasion and metastasis. Loss of basement membrane continuity is a hallmark of the invasive process. Once the basement membrane is damaged, malignant cells gain access to the vascular/lymphatic compartment of the microcirculation through the subendothelial basement membrane. This phase of the invasive process is facilitated by the rich tumor vascular supply encouraged by tumor-induced angiogenesis, as well as the presence of various defects in the basement membranes of the newly formed vessels (Bass et al. 2009). In recent years, research has highlighted that tumorigenesis is dependent on a small subset of cells within the tumor, called cancer stem cells. There is evidence that abnormal expression of the transcription factors Oct4, SOX2, and Nanog are involved in the carcinogenesis of various cancers (Kokko et al. 2011;Todoroki et al. 2016). Reduced SOX2 expression is associated with loss of the pluripotent state and a propensity for differentiation, highlighting a correlation between SOX2 and various malignant tumor types. SOX2 is abundant in 41% of lung small cell tumors and 43% of basal cell carcinomas of the breast. Conversely, reduced expression of SOX2 has been described in some types of carcinoma, such as gastric cancer and choriocarcinoma (Shi et al. 2012). Bass et al. (2009) demonstrated mutation of the SOX2 gene transcription factor in esophageal and lung SCC, which led to increased proliferation and differentiation of basal cells. In turn, Jing and Zheng (2010) highlighted the expression of SOX2 in cervical cancer and observed increased proliferation, clonogenicity, and tumorigenesis both in vivo and in vitro. There is evidence that SOX2 promotes the growth of SCC tumors in vivo through the activation of AKT/signaling mTORC1, which promotes cell proliferation (Bass et al. 2009). The oncogenic performance of squamous cell carcinoma of various organs, such as the esophagus, is associated with expression of SOX2, predominantly in basal progenitor cells of the stratified epithelium.
These findings indicate that basal stem/progenitor cells are the origin of cells of SCC; further research indicates that cooperation between SOX2 and microenvironment-activated STAT3 is required to drive SOX2-associated carcinogenesis (Watanabe et al. 2014;Wang et al. 2018). STAT3 (when activated by phosphorylation) modulates cellular biological behavior, including cell cycle progression, epithelial-mesenchymal transition, inflammatory response, and angiogenesis. The expression of p-STAT3 in tumor samples has been associated with the negative prognosis of several tumors, such as breast cancer, lung cancer, and lymphoma. However, the exact prognostic role of p-STAT3 in penile cancer remains unclear. Increased expression of p-STAT3 is associated with unfavorable outcomes in patients with colorectal cancer, gastric cancer, hepatocellular carcinoma, esophageal cancer, and pancreatic cancer. In addition, other molecules involved in carcinogenesis, such as the cell surface glycoprotein (CD44), promote cell-cell interaction, adhesion, and cell migration. Oh et al. (2013) showed that CD44 is a marker for head and neck tumor squamous cell carcinoma, and that tumor stem cells expressing CD44 have an increased ability of initiating carcinogenesis. This was revealed by an immunohistochemical study in that the overexpression of CD44 can be a sign of cellular aggression and a negative prognostic factor of pharyngeal and laryngeal SCC, which helps clinicians make therapeutic decisions. We evaluated tumor performance in association with the expression of biomarkers SOX2, STAT3, and CD44 high /CD24 low in the tumor tissue by flow cytometry.

Study population
Thirty-four men, over 18 years of age, with confirmed clinical, laboratory, and staging diagnosis for penile tumor were evaluated in an observational, prospective, and translational study. Additionally, a control group from the same service provider where the study was conducted was evaluated.

Patients and controls uptake procedures
The patients included in the study had histological analysis of the tumor biopsy performed by two pathologists. The collection of clinical data for the study was performed using standardized clinical records. When necessary, the medical records were searched to obtain additional clinical information.
The control group comprised male volunteers, accompanying patients and employees of the service where the study was conducted who, after clinical evaluation to select the eligible participants, followed the inclusion criteria for the controls. These were absence of prolonged use of immunosuppressants, absence of clinical history of infection by human immunodeficiency virus 1 and 2, and human lymphotropic virus 1 and 2; absence of clinical history of infection in the last 30 days and at the time of peripheral blood collection; absence of clinical history of viral hepatitis and autoimmune disease; and absence of previous history of any chronic disease.

Biological sample collection
Tumor tissue was collected during the surgical procedure, and two small tumor fragments were removed to conduct this study. The fragments were stored in a sterile flask containing Roswell Park Memorial Institute (RPMI) cell culture medium for analysis of CD44 high /CD24 low , SOX2, and STAT3 expression in the tissue by flow cytometry.

Enzymatic digestion of tumor tissue
For analysis of the tumor tissue, enzymatic digestion was performed to remove the stroma and release the cells present in the tumor microenvironment. The tissue samples collected in RPMI medium supplemented with penicillin-streptomycin (Sigma-Aldrich, USA) were subjected to enzymatic digestion (dissociation) with collagenase I (Gibco, USA) to obtain cell suspension, according to the manufacturer's instructions. The tissue was divided into small fragments (2 mm) with the aid of a sterile scalpel and immersed in the collagenase I solution at a concentration of 2 mg/mL in Hanks' Balanced Salt Solution. These were then incubated in a dry bath at 37 °C for 2 h under agitation. After incubation, 2 mL of phosphate-buffered saline (PBS, pH 7.4) was added and filtered through a 35 µm filter (BD Falcon, BD Biosciences, USA) and the cell suspension was centrifuged at 250×g for 5 min. Subsequently, the supernatant was discarded and the cell precipitate was washed with PBS and centrifuged at 250×g for 5 min. Then, the cell precipitate was resuspended in 1 mL of PBS and 100 µL was removed to perform the protein analysis by flow cytometry.

Flow cytometry technique
Flow cytometry is an advanced and rapid technology that allows counting, examining and sorting of individual cells and measures multiple parameters simultaneously, providing a quantitative analysis. In the flow cytometer, the fluid system channels a sample of cells in a single stream such that the cells pass one by one through a focused beam of a laser. As each cell passes through the beam, the light scattering properties and fluorescence are collected by the optical system and directed to various detectors. The signals received by the detectors are then converted into numerical values by the electronics and analyzed in specific software. As a cell passes through a laser beam, light is scattered in different directions and two parameters are defined: cell size, represented by forward scatter, and cell complexity (granularity), represented by side scatter. This information is visually displayed in a dot plot, in which each dot represents a single cell, differentiated by the way it scatters light and by fluorescence, which is proportionally direct to the quantity of molecules expressed in the cell.

Immunophenotyping by flow cytometry
The first step was the analysis of proteins expressed on the cell surface (CD44 and CD24). Cells were fixed with 100 µL of pre-warmed BD Cytofix/Cytoperm™ and incubated for 10 min at 37 °C. Then, after centrifugation at 250×g for 8 min, 5 µL of monoclonal antibodies, anti-CD44 and anti-CD24 (BD Biosciences, USA), were added. Monoclonal antibodies are conjugated with fluorophores, such as PE, FITC, PerCP, and PECY-7, which differ from each other with respect to the color they emit. Monoclonal antibodies have specificity for molecules expressed on the membrane and inside cells, which makes it possible to characterize human leukocytes and other human cells and to analyze their functions. These molecules are identified as "clusters of differentiation" (CD).
STAT3 and SOX2 proteins are expressed inside the cells (cytoplasm and nucleus, respectively). For detection of these proteins, cell permeabilization was performed using BD™ Phosflow reagents (BD Biosciences, USA). After incubation for 20 min at room temperature and protected from light, cells were permeabilized with 750 µL of Perm Buffer III (BD Biosciences, USA) at 4 °C and incubated for 30 min submerged in ice, protected from light. The cells were then washed with 3 mL of PBS and centrifuged at 250×g for 8 min, and the supernatant was discarded. Three washes with PBS were performed. Subsequently, intracellular labeling was performed with the monoclonal anti-SOX2 and anti-STAT3 antibodies and incubated for 40 min protected from light. Then, a new washing step was performed, the supernatant was discarded and the cells were resuspended in 300 µL of staining buffer and acquisition was performed in the flow cytometer. After the washes, acquisition of 50,000 cellular events was performed using the FACSVERSE equipment (Becton Dickinson, Sunnyvale, CA) and the analyses were performed in the FACSSUÍTE program (Becton Dickinson, Sunnyvale, CA) and expressed in percentage values.

Statistical analyses
The Shapiro-Wilk normality test was applied to the numerical variables. Those that presented a normal distribution were analyzed by calculating the mean and standard deviation (SD), while the median was calculated for the variables that did not present a normal distribution. The Mann-Whitney U test was used to compare numerical variables between groups with non-normal distribution.
A statistical significance level of p < 0.05 was adopted. All statistical analyses were performed using GraphPad-Prism v6.0 (GraphPad Software, San Diego, CA).

Perineural invasion
We assessed a total of 27 patients with SCC of the penis who presented with presence (IP+) or absence (IP−) of perineural invasion for tumor expression of CD44, CD24, SOX2, and STAT3. No significant difference was observed in the percentage of CD44, CD24, SOX2, and STAT3 expression between the IP+ and IP− groups (Fig. 1). However, we found a higher median percentage of STAT3 expression in the IP− group when compared with the IP+ group, as shown in Fig. 2a.
Evaluation of markers CD44, CD24, SOX2, and STAT3 in tumor tissue of patients treated with lymphadenectomy. Patients who underwent lymphadenectomy were divided into those with positive lymphadenectomy (L+) and negative lymphadenectomy (L−). For the analysis of the CD44 percentage, 27 patients were examined: 10 patients in the L+ group and 17 in the L− group. The graph of the second line with the first column of Fig. 3 shows that there are patients with discrepant percentages of CD44 in both groups; however, the groups do not present evidence of decentrality, since the distance between the mean and median values in both groups is not large (L+: mean = 49.51, median = 43.65; L−: mean = 37.51, median = 32.00).
In the CD44 immunofluorescence analysis, 26 patients were evaluated (10 with L+ and 16 with L−). The L+ group shows evidence of decentralization, as the mean is more than twice the value of the median (mean = 26,717, median = 11,290). The graph of the second line and second column shows that some patients in this group had above-mean CD44 production, which possibly contributed to the mean increase. The L− group had patients with a concentration of CD44 production around the mean value, and although some patients displayed CD44 production well below the mean, there was no evidence of decentralization in this group. The difference between the means of the two groups was significant. We conclude that a significant difference in the mean CD44 expression exists between the two groups.
In the context of C24 percentage, 16 patients were examined (5 with L+ and 11 with L−). The group of patients with L+ displayed a slight decentralization, demonstrated in the graph of the first line with the first column, although a difficulty in the observation of this group was found because of the small number of patients. In the L− group, some patients with high percentages of CD24 expression probably contributed to the group mean increase. The difference between the means of the two groups was statistically significant. However, no significant difference was detected between the mean percentages of CD24 of the two groups. Neither group displayed evidence of decentralization concerning CD24 immunofluorescence. The L+ group showed values considerably closer to the mean; the L− group showed one patient with a value well below the mean, with the highest concentration of values present around it. This is presented in the graph in the second row, first column of Fig. 3.
A total of 24 patients were observed for Sox2 percentage (10 with L+ and 14 with L−). There were several patients with discrepant Sox2 percentages in both groups, as displayed in the graph in the third row, first column of Fig. 3, although there was evidence of decentrality only in the L+ group, since the mean value of Sox2 percentage deviates considerably from the median value (mean = 28.66, median = 8.14) in this group. Considering the immunofluorescence of Sox2, decentralization of the data was detected only in the L+ group, as shown in the graph of the third row and second column of Fig. 3, because, only in this group did the discrepant points appear to be influencing the increase of mean Sox2 immunofluorescence (mean = 12,536, median = 4789).
Regarding STAT3 percentage, 16 patients were evaluated (5 with L+ and 11 with L−). In the group with L+, the existence of patients with extreme percentages, both higher and lower than the mean, balanced the group (see graph of the fourth row-first column of Fig. 3), but caused an increase in group variability (SD = 24.10). Therefore, no evidence of decentralization in this group was detected. The same occurred in the L− group, with extreme values above and below the mean value that appear to stabilize the mean itself; therefore, again, no significant difference between the mean and median of the group was present, as shown in the graph of the fourth line first column of Fig. 3.
In the evaluation of the STAT3 immunofluorescence, the group with L+ did not present with decentrality, as displayed in the graph, even with the presence of patients with a STAT3 immunofluorescence slightly distant from the mean. Although there are patients with STAT3 immunofluorescence significantly below the mean in the L− group, as represented in the graph of the fourth line, second column, there was no evidence of decentralization, because the mean of the STAT3 production of the group was not far from the median (mean = 920.9, median = 876.0).
Evaluation of the markers CD44, CD24, SOX2 and STAT3 in the tumoral tissue of patients with penile SCC, in the context of age. Patients were divided into two groups based on age. The first group refers to patients aged younger than 60 years, and the second group refers to patients aged 60 years and older. Among patients with values for CD44 percentage (CD44%), 18 patients were younger than 60 years and 16 patients were older than 60 years. Patients aged younger than 60 years had a median of 32.88 and a mean of 42.56 for CD44%, which shows data decentrality. The same occurred in the group of patients older than 60 years, where the median value was 28.28 and the mean was 34.56. Thus, there was no evidence to reject the hypothesis that the means in both groups were similar.
Concerning the CD24 percentage (CD24%), 23 patients were examined. Eleven patients younger than 60 and 12 patients older than 60 were present. Therefore, the data were not central, and there was great heterogeneity in the group. The graph of the first column of the second row of Fig. 4 shows that both groups have outliers, although the group of patients younger than 60 showed better distribution around the mean. The appropriate test for the difference in means had a p-value close to 0.05. Thus, it can be considered that there is a significant difference between the CD24% mean for patients younger than 60 years and that for patients older than 60 years.
For the Sox2 percentage (Sox2%), 30 patients were examined; 16 were classified younger than 60 years, and 14 as older than 60 years. The third row of the graph of the first column of Fig. 4 shows a large number of outliers in both groups. In the group under the age of 60, the mean (28.25) was shown to be much higher than the median 16.35), because of which there was a decentralization in the values of the percentage of Sox2 in this group. In the group older than 60, the mean was more than double the median, revealing heavy decentralization of data. In both groups, the SD value exceeds the mean, because of which both groups were very heterogeneous.
Regarding the STAT3 percentage (STAT3%), 23 patients were evaluated, 11 younger and 12 older than 60 years of age. This is displayed in the chart of the fourth row of the first column of Fig. 4.
Evaluation in tumor tissue of CD44, CD24, SOX2 and STAT3 markers in relation to the presence of angiolymphatic vascular invasion.
The variable vascular invasion was divided into two groups: a group with positive vascular invasion (IV+) and another with negative vascular invasion (IV−).
For the percentage of CD44, 34 patients were examined, where 11 patients were categorized into the IV+ group and 23 into the IV− group. In the group with IV+, the mean (26.81), was not distant from the median (21.12), although the percentage of CD24 expression was not well distributed around the mean, due to some discrepant values above and below the mean, demonstrated in the graph of the first line and first column of Fig. 5. In the group with IV−, the number of patients with a high percentage of CD24 is shown to be influencing the decentralization of the data (graph of the first line and first column of Fig. 5.), since the mean (44.91) was above the median (34.23). The test for the difference of the mean did not achieve a significant value.
Upon CD44 immunofluorescence analysis, the group with IV+ did not show signs of decentralization, since the mean (8412) CD44 production was not far from the median (8587), and most patients displayed a CD44 production value close to the mean value of the group, as shown in the graph of the first row second column. In the group with IV−, some patients with CD24 production above the mean value caused a decentralization in the group (graph of the first row, second column).
Regarding of CD24, 25 patients were examined, 9 patients with IV+ and 16 with IV−. In the group with IV+, most patients obtained a percentage of CD24 well below the mean, as shown in the graph of the second line, first column; however, patients with high percentages caused decentralization of the group since the mean value was three times the median value.
Of the 23 patients examined for immunofluorescence analysis of CD24, 9 belonged to the IV+ group and 14 belonged to the IV− group. Even with the presence of some discrepant points, the IV− group did not display evidence of decentralization, as shown in the graph of the second line, second column. In the group with IR+, most patients had a CD24 production below the mean, and patients with high CD24 production values contributed to both an increase of the mean and a decentralization in the group (graph of the second line, second column).
To evaluate the percentage of Sox2, 30 patients were examined, 10 with IR+ and 20 with IR−. Both groups displayed discrepant values, causing a decentralization of the data, shown in the graph of the third line, first column of Fig. 5. In the group with IR+, most patients had a percentage of Sox2 close to the mean. In the group with IR−, there was a small concentration of patients with percentages close to the mean, although the high values were influencing the growth of the mean (see graph of the third line first column of Fig. 5).
Although the two groups seem to be well distributed around the mean in the context of Sox2 immunofluorescence analysis, as shown in the graph of the third line, second column of Fig. 5, the distances between the mean and median in both groups showed that the production of Sox2 in some patients caused an increase in the mean (IV+: mean = 8056, median = 5100; IV−: mean = 11,409, median = 6638). Therefore, there is evidence of decentralization in both groups.
For the STAT3 percentage analysis, 23 patients were examined, 9 with IV+ and 14 with IV−. In the group with IV−, 50% of the patients were concentrated below the mean and 50% are concentrated above; no evidence of decentralization in this group was detected, although some patients displayed higher STAT3 percentages in Fig. 4 Expression of CD24+, CD44+, SOX2, and STAT3 in percentage of cells and immunofluorescence in patients with SCC of the penis according to age comparison to the other patients in the group (graph in fourth line, first column of Fig. 5). The appropriate test for the mean difference between the two groups indicated a p-value of 0.0095. A difference in the STAT3 percentage mean of the two groups was detected. In the STAT3 immunofluorescence analysis, both groups presented discrepant points, as shown in the graph of the fourth line (second column of Fig. 5), although, in both groups, there was no evidence of decentrality. The means of STAT3 production in the groups were significantly different (p = 0.0183).

Discussion
In this study, patients with penile cancer seen at Hospital de Câncer de Pernambuco (HCP) were evaluated. This is the first study developed at HCP that shows the expression profile of CD44 high CD24 low , STAT3+, SOX2+ and TCD3+ in the tumor tissue of penile cancer. The HCP is responsible for the care of 55% of cancer patients in the state of Pernambuco, is a hospital that exclusively serves exclusively users of the Unified Health System (SUS). We Fig. 5 Expression of CD24+, CD44+, SOX2, and STAT3 in the context of cell percentage and immunofluorescence in patients with SCC of the penis presenting with angiolymphatic vascular invasion evaluated 38 patients with penial cancer with a median age of 61 years (37-80 years). Globally, the age range of penile cancer patients is observed as ranging between 40 and 70 years. In Brazil the average range is between 55 and 60 years, while in developed countries the average tends to be higher, around 70 years (Favorito et al. 2008). Hernandez et al. (2008) showed that in the USA there were more deaths and diagnoses at earlier ages among people of African descent, perhaps due to the fact that this group, even in the USA, is mostly of low socioeconomic status.
The classification of the National Comprehensive Cancer Network (VanderWalde et al. 2016), considers as young elderly, those patients aged between older than 65 and younger than or equal to 75 years; elderly, between older than 76 and younger than or equal to 85 years; and very elderly, older than 85 years (Balducci 2006). The WHO in 2015 considered in developing countries, elderly, people over the age of 60 years, according to the National Policy of the Elderly (PNI), provided by Law No. 8. 842, January 4, 1994, being the same applied by the Statute of the Elderly, Law No. 10.741, October 1, 2003(BRASIL 2004. Based on the WHO classification, the patients in this study were distributed into two groups aged ≤ 60 and > 60 years. In the description of clinical characteristics, the glans was the most affected with the primary tumor with (76%) of cases, followed by the penile shaft (14%). Wanick et al. (2011) showed a series of SCC of penis with 91% of this sample on glans. The highest frequency of pT2 stage (55.3%) and 5 cases of metastatic disease (13.2%). Immunotherapy has become increasingly important in the treatment of cancer, and an understanding of the mechanism of immune relationships in the tumor microenvironment is necessary. The importance of tumor-infiltrating B cells has not been highlighted in previously published literature; however, recently, a significant influence of B cells on tumor growth has been described. Consequently, this cell population is now also perceived as a possible therapeutic target. Understanding the immunological relationships of these cells and tumor microenvironments may help to effectively combine standard approaches, including surgery or radiochemotherapy with immunotherapy (Schuler et al. 2018).
In the series under evaluation, patients with penile cancer showed reduced percentage values of B lymphocytes (p = 0.0006) when compared with healthy controls (Fig. 1a). The patients in our sample showed percentage values of total T (p = 0.81), TCD4+ (p = 0.15), TCD8+ (p = 0.40), and NK cells (p = 0.16) similar to those of healthy controls, although there is growing evidence that immune cells can trigger various mechanisms that increase tumor growth and metastasis; however, no data provide conclusions regarding peri-tumor inflammation for penile cancer. Vassallo et al. (2015) evaluated the presence of CD3, CD4, and CD8 by immunohistochemical analysis. In patients aged 60 years and above, presence of lymph node metastasis, urethral invasion, and high histological grading, multivariate analysis reported an unfavorable outcome. Improved 5-year disease-free survival was significantly associated with disease stage I + II, N0 lymphatic status, and absence of perineural, vascular, or urethral invasion on multivariate analysis. Presence of lymph node metastasis, perineural and vascular invasion, and low inflammatory infiltrate displayed a significantly less favorable outcome (Vassallo et al. 2015).
Perineural invasion was defined as tumor cells in contact with or invading a nerve. The perineural space is a recognized route for SCC invasion. The presence of perineural invasion is associated with loco-regional recurrence and decreased survival in patients with head and neck SCC. The positive associations between perineural invasion and local-regional and distant recurrence, survival, cervical lymph node metastasis, and dissemination were confirmed. In this series, perineural invasion was present in more than half of the cases (52%) and it was associated with local recurrence and disease-specific mortality. The association between perineural invasion and local recurrence was independent of the tumor stage; therefore, this tumor behavior suggests that tumors that invade the perineural space are biologically more aggressive. SCC with perineural invasion presents with a higher possibility of local recurrence in head and neck SCC. Perineural invasion is considered a negative independent predictor of survival.
We analyzed the expression of CD44, CD24, SOX2, and STAT3 biomarkers in our sample's tumor tissue in the context of perineural invasion (IP+) or not (IP−). No significant difference was observed in the percentage of expression of CD44, CD24, SOX2, or STAT3 between the IP+ and IP− groups (Fig. 1). However, we found a high median in the percentage of STAT3 expression in the IP− group when compared with the IP+ group, as shown in Fig. 2A.
Regarding the percentage of CD44 count (CD44%), 27 patients were examined. Group CD44% means were not significantly different. However, this could be due to small number of patients, as the sample is half the size needed for accurate evaluation. In the evaluation of CD44 expression by immunofluorescence, the mean and median values showed an important numerical difference, but with a high value of the SD; thus, this finding should be reviewed. The test for the difference of means between the groups did not show a significant p-value.
With respect to the percentage of CD24, the sample size was small and the difference between the mean and median values was minimal. Theoretically, this group should present a more aggressive behavior of the neoplasm, although the difference between the mean and median was not as small in the IP− group as that in the IP+ group. The data are not suggestive of a change in behavior. However, this finding can be justified by the presence of a patient with an outlier value in the IP− group, which contributes to the slight increase in the mean in this group. No significant difference between the mean CD24 percentages of the two groups was found. Regarding the analysis of CD24 immunofluorescence expression, both groups had values very close to the mean, with no significant change found.
The evaluation of Sox2 percentage was performed in 24 patients. In the IP− group, there was a considerable number of patients with Sox2 percentages close to zero. This means that increases in Sox2 percentage may be associated with a worse prognosis; however, this hypothesis should be verified using a sample with a larger number of patients and less heterogeneity. Regarding Sox2 immunofluorescence expression, a significant difference was detected between the means of Sox2 immunofluorescence in the two groups (p = 0.034). This difference is also associated with the Sox2 percentage, corroborating the possibility that overexpressed Sox2 is associated with a worse prognosis.
In the evaluation of STAT3 percentage despite the small number of 16 patients, a below-mean STAT3 concentration was detected in the IP− group. In the IP+ group, three patients displayed a higher STAT3 percentage than the other four; for this reason, the mean was far from the median. Therefore, a significant difference between the means of STAT3 percentage of the two groups cannot be inferred. This observation must be verified using analyses including samples with larger sample sizes. In the evaluation of the STAT3 expression by immunofluorescence, no differences were observed between these groups. The number of patients in the sample evaluated for STAT3 immunofluorescence expression was greater than the sample evaluated for the percentage, which reduces the quality of conclusions drawn from the analysis.
In SCC of the penis, lymphadenectomy is part of neoplasm treatment and staging. We evaluated the expression of CD44, CD 24, SOX2, and STAT3 in the tumor tissue of patients who were treated with lymphadenectomy. We divided the patients into two groups: those with indications to undergo lymphadenectomy (L+), and those with no indication to undergo lymphadenectomy (L−). A significant difference between the mean CD44 of the two groups was detected (p = 0.038). This first analysis corroborates a behavior similar to head and neck SCC, where the lowest CD44 expression is associated with lymph node involvement.
In the evaluation of the percentage of CD24, the first difficulty in studying this group is the small number of 16 patients. The group that was treated with lymphadenectomy showed small differences, although the group of patients that did not undergo surgery showed high CD24 percentages, which probably increased the group's mean value. The p-value of the test of difference of means between the two groups was significant (p = 0.023). The evaluation of CD24 immunofluorescence expression in both groups did not present with number discrepancies. The values found in the group of patients who received lymphadenectomy were close to the mean; this was similar in the group that did not undergo lymphadenectomy, despite one patient with a value well below the mean.
In the evaluation of the data of Sox2 percentage in the context of lymphadenectomy (n = 24), several patients with Sox2 percentages distant from the mean in both groups were present, although in the group treated with lymphadenectomy, a mean of the percentage of Sox2 was far from the median. Despite the early phase of the study, where the number of patients was still small, this finding may indicate that Sox2 percentage may be a prognostic marker. Sox2 immunofluorescence expression values in the L+ group displayed a mean distinctly different from the median, but with points of discrepancy interfering in the result, producing a higher mean.
The STAT3 percentage analysis was performed in 16 patients not producing a significant difference between the mean and median of the group. The quality of the initial analysis may have been reduced by the small number studied.
We divided patients into two age groups. The first group refers to patients younger than 60 years, and the second group refers to patients 60 years and older. Perhaps this is not the best division because of group heterogeneity, which may cause distortions in the results of the analysis. The data showed a mean of 42.56 and a median of 32.88 in patients under 60 years of age; in the group over 60 years, similar behavior was observed, where the mean was 34.56 and the median, 28.28. This fact shows the non-centrality of the data, because the two groups had discrepant points. A small number of patients have a CD44 percentage much higher than others in both groups. In the same way, in the analysis of the CD44 immunofluorescence expression (CD4 MFI), the group of patients under 60 years of age had a mean greater than twice the median, making non-centrality of the data evident, and leading us to conclude that the production of CD44 was much higher in some patients than in others. The same finding (significant difference between mean and median) was observed in the group of patients aged over 60 years. By evaluating the SD of the two groups, it is possible to infer that there is great heterogeneity within each group, since the SD is higher than the mean.
In the evaluation of the CD24 marker in the group of patients aged under 60 years, there was no great difference between mean and median, although in the group of patients aged over 60 years, the mean was almost twice the median, and the SD was even greater than the mean. The data found so far on CD24 immunofluorescence expression (CD24 MFI) showed a centralization of the numbers for the group under 60 years of age, while in the group over 60 years of age, there was a concentration of values around the mean, as well as some discrepant values that caused the mean of CD24 MFI to increase. An increase in sample size may correct this discrepancy, confirming or refuting the significance of our findings.
Evaluation of the data regarding the percentage of Sox2 (Sox2%) was conducted. The heterogeneity of the groups may compromise the final analysis; perhaps the analysis of a larger number of patients will help us achieve balance. The data of Sox2 immunofluorescence expression (Sox2 MFI) were very similar to those of Sox2%, with a large distance between the mean and median. This may mean that in patients with penile cancer, age does not interfere percentage or with Sox2 immunofluorescence expression.
When the STAT3 percentage was evaluated, the difference between the mean and median in both groups were not large, although when the SD was analyzed, a high value obtained showed that there was heterogeneity in both age groups. The STAT3 immunofluorescence expression (STAT3 MFI) had discrepant values in the group of patients under 60 years of age, although in the group of patients above 60 years of age, the mean and median were very close, and the discrepant points did not seem to be affecting the mean value. Once more, observation may be compromised by the number of cases evaluated, and an increase in the number of patients is expected to resolve this issue.
Vascular invasion, when present, may be related to a greater possibility of metastasis in SCC of the penis. Thus, the invasion variable was evaluated in the context of expression of CD44, CD24, Sox2, and STAT3, dividing the sample in two groups: a group with positive vascular invasion (IV+) and a group with negative vascular invasion (IV−).
In evaluating the percentage of CD44, data from 34 patients were examined. Patients who presented with IV+ had mean and median values that were similar, although the comparison with those with IV− was not statistically significant. In the evaluation of CD44 expression by immunofluorescence, the group with IR+ did not present a significant distance from the numbers that represent the mean and median. In the group of patients with IV−, some patients displayed values of CD44 production far above the mean value, which may have increased the distance between the mean and median in the group.
Regarding the study of the percentage of CD24, in the group with IR+, most patients had a CD24 percentage well below the mean; however, the presence of patients with high percentages may be reducing the quality of observation in the group, with the mean value being three times greater than the median value. In the group without vascular invasion, the patients with data far from the mean did not change the mean value. This sample behavior must be related to the number of cases being analyzed at the moment; an increase in the sample size would correct the distortion. To evaluate the expression of CD24 by immunofluorescence, the presence of distant points from the median in group IV− does not apparently alter the results. In the group where the presence of vascular invasion was registered, there was a higher number of patients with lower CD24 production, and a few with a high production, which, at the moment, contributes to the increase of the mean expression; this observation leaves doubts about the behavior of CD24 expression in SCC of the penis.
For Sox2 percentage, we used data from 30 patients, who were distributed into two groups (10 patients with vascular invasion [IV+] and 20 without vascular invasion [IV−]). In both groups, the presence of Sox2 percentage values far above and far below the mean changes its value. In the positive group, most patients had a Sox2 percentage expression close to the mean value; however, in the group without vascular invasion, we found that the sample was contaminated by high-value outliers, which influenced the results. When both groups were studied in relation to the Sox2 immunofluorescence expression, the difference between the mean and median in both groups shows that the Sox2 expression of some patients influences the mean result.
The data of the STAT3 percentage were evaluated in 23 patients who were divided into two groups, one of 9 presenting with vascular invasion (IV+) and another of 14 without vascular invasion (IV). In group IV−, half of the patients were below the mean and the other half were above it, although there were patients with high STAT3 percentages when compared with others in the same group. In the group with vascular invasion by the tumor, the high percentage values of some cases influenced the mean. Thus, there was a significant difference between the mean of STAT3 percentage of both groups (p = 0.0095). Confirmation of this result upon analyzing a larger sample can affirm STAT3 as a useful prognostic marker. When the STAT3 immunofluorescence expression was evaluated, both groups presented points far from the mean; however, despite this influence, the calculation of a significant p-value (p = 0.0183) leads us to believe that STAT3 should be further investigated as a possible prognostic marker.

Conclusion
Our study indicates that it may be possible to characterize a molecular marker in patients with SCC of the penis that accurately forecasts case prognosis, since penile SCC treatment can potentially be highly morbid, and determining the aggression of treatment is essential. Such a marker will make it possible to distinguish patients who will benefit from a more aggressive treatment from those who will not, depending on the pathology analyzed. Moreover, our research group will continue this study to obtain larger samples with more